CN109002747A - Visible remote sensing image Ship Detection on star based on geometrical characteristic - Google Patents
Visible remote sensing image Ship Detection on star based on geometrical characteristic Download PDFInfo
- Publication number
- CN109002747A CN109002747A CN201710421783.4A CN201710421783A CN109002747A CN 109002747 A CN109002747 A CN 109002747A CN 201710421783 A CN201710421783 A CN 201710421783A CN 109002747 A CN109002747 A CN 109002747A
- Authority
- CN
- China
- Prior art keywords
- slice
- mass center
- point
- potential
- ship target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
This application discloses a kind of naval vessel confirmation methods based on the naval vessel plane geometry characteristic of visible remote sensing image on star, mainly include following part: binarization operation being carried out to slice first with the mean value and standard deviation of ship images undetermined slice, and finds out the centroid position of the potential target after binaryzation in slice;Then potential target and background in slice are isolated out using relative entropy according to the centroid position that acquires and updates binarization result and the center of mass point position of slice;According to the length-width ratio of slice, point situation finds out the center of mass point of slice sub-block at a distance from potential target mass center to centre of slice, and the symmetry axis of potential target is found out according to the center of mass point of potential target center of mass point and sub-block;Finally using obtained symmetry axis find out potential target length-width ratio and potential target where circumscribed rectangular area, determined with the two values, obtain the confirmation result of final ship images undetermined slice.
Description
Technical field
The application belongs to digital image processing techniques field, is related to a kind of goal verification method in remote sensing images, especially relates to
And it is a kind of on satellite visible remote sensing image Ship Target detection method.
Background technique
China sea area is wide, and the marine territory area possessed reaches 299.7 ten thousand square kilometres, and naval vessel is as seaborne
Important carrier has a wide range of applications in each field.Therefore, realize the monitoring on naval vessel to the maritime rights and interests of maintenance this country of China,
Supervising territorial waters exclusive economic zone has very great meaning.At civilian aspect, realize that naval vessel detection can be each with real time inspection
The shipping traffic situation at harbour, fast searching vessel in distress monitor illegal fishing and smuggling boat;Militarily, it can monitor
National sea area safe condition and investigation enemy sea military strength and deployment, provide safeguard for national security and territorial integrity.
And naval vessel detection is realized on star can reduce the burden that a large amount of satellite image output transmissions come and reduce data transmission bandwidth
Waste, the real-time for improving naval vessel monitoring.So there is very important grind based on visible remote sensing image naval vessel detection on star
Study carefully value and application prospect.It directly influences the real-time and efficiency of the monitoring of ocean Ship Target, while removing invalid sea
Foreign image data, improve data transfer efficiency.
The method for carrying out naval vessel confirmation to visible remote sensing image at present can be divided into two major classes: (1) traditional characteristic extracts
Method;(2) based on the method for machine learning.The former is based primarily upon human visual system and obtains or from known naval vessel sample
The ship images feature that the naval vessel prior information of the gray scale, the texture that count etc. obtains, for example, wavelet character, based on sea analyze
Anomaly extracting, naval vessel histograms of oriented gradients feature etc., such methods have higher for the naval vessel in certain special scenes
Detection accuracy, however for different scenes and do not have universality;The latter mainly utilizes naval vessel sample set to certain nerve
Network frame is trained to obtain naval vessel judgment models, more representational mainly to have support vector machines and convolutional neural networks
Deng, such methods have stronger universality, but calculate complexity, and the quantity of floating number is more in the model for needing to save, and by
In the parameter evidence, there are irradiation space, being easily damaged on star in model, will lead to model can not normal use, therefore such method is simultaneously
It is not suitable for carrying out naval vessel detection on satellite.Although there are many method of naval vessel detection, there are no can cope with warship on star so far
The efficient algorithm that ship detection application harsh memory space and calculation amount require.
In this context, storage and calculating in the accuracy and star that naval vessel detects on visible remote sensing image in order to balance
Harsh conditions need to analyse in depth the Ship Imaging feature in visible remote sensing image, study a kind of true based on the naval vessel on star
Recognizing algorithm is particularly important.
Summary of the invention
The application technical problems to be solved are to provide the visible remote sensing image ocean Ship Target that can be used on star
Confirmation method.Imaging features information of this method according to Ship Target on visible remote sensing image, by the gray scale and line on naval vessel
The description of reason feature, shape feature are combined with multiple technologies such as the naval vessel for the naval vessel plane characteristic for adapting to different background confirmations, are had
Effect improves the accuracy rate that naval vessel detects in visible remote sensing image, and reduces computational complexity and the requirement to storage.
In order to realize that above-mentioned goal of the invention, the application use following technical solutions:
Naval vessel confirmation method based on the naval vessel plane geometry characteristic of visible remote sensing image on star, it is characterised in that including
Following steps:
(1) it extracts multiple ship images to be detected to the visible remote sensing image of input to be sliced, wherein each is to be checked
Surveying in ship images slice only includes a potential Ship Target;
(2) ship images to be detected slice is traversed, finds out the mean value and standard of grey scale pixel value in the slice
Difference, and the binary map g' for carrying out binarization operation to obtain the slice is sliced to described image, it is asked according to the binary map g'
Obtain the mass center m for the potential Ship Target T' for including in current slice0' position;
(3) using obtained mass center m in above-mentioned steps (2)0' position and binary map g', search mass center m0' belonging to connect
Lead to region and according to each pixel and mass center m0' relative entropy as a result, accurately approaching the side in the region by K-means algorithm
Boundary, and updated binary map g is obtained, potential Ship Target T and its mass center m0;
(4) the length-width ratio r and the mass center m of the slice are calculated0To the distance of the centre of slice point, according to naval vessel shape
Shape characteristic obtains the sub-block with symmetric relation from ship images to be detected slice, and calculates the mass center m of the sub-block1Position
It sets;
(5) according to above-mentioned two mass center m0And m1Position, calculate the symmetry axis straight line of potential Ship Target T, and obtain
Locality of the potential Ship Target T in the slice;
(6) with the mass center m in the slice0For starting point, the symmetry axis straight line obtained according to above-mentioned steps (5) exists respectively
Symmetry axis straight line two sides, which are found, belongs to potential Ship Target T, and the point e farthest apart from symmetry axis vertical range1And e2, and find
The point v that the symmetry axis intersects with potential Ship Target T external periphery outline1And v2;
(7) the area M of potential Ship Target T minimum circumscribed rectangle, and the point e according to obtained in step (6) are calculated1, e2,
v1And v2Each point is calculated separately to mass center m0Distance ed1, ed2, vd1And vd1, according to described apart from ratio calculatedIf the area M and ratio rlsMeet default Rule of judgment, then determine that potential Ship Target T is naval vessel,
Otherwise determine that potential Ship Target T is not naval vessel.
Wherein, in the step (2), the threshold value that when binarization operation uses is the sum of mean value and standard deviation.It is described undetermined
Ship images slice be obtained after tentatively extracting it is uncertain whether be Ship Target image slice.
In the step (3), the lookup of connected region uses 8 fields;Its each gray scale first is counted to the slice
Be worth probability of occurrence, then by outside mass center and mass center region it is random a little centered on, calculate relative entropy with by the slice
It is divided into two class of potential target region and background area.
Naval vessel confirmation method provided by the present invention based on the naval vessel plane geometry characteristic of visible remote sensing image on star
It has the advantages that
1. calculating, computation complexity used when naval vessel plane geometry feature is low, calculation amount is small, the number of required preservation
According to naval vessel confirmation that is few, being detected suitable for naval vessel on star.
2. current pixel point is measured at a distance from cluster centre using relative entropy in algorithm, when accurately approaching boundary
There is preferable segmentation effect, potential target and background in slice can be more accurately distinguished between open.
3. the plane geometry feature on naval vessel has universality, the variation under various illumination and marine background is smaller, can fit
For correctly differentiating Ship Target and non-ship target in the remote sensing images under marine environment and different resolution complicated and changeable.
Detailed description of the invention
The application is described in further detail with reference to the accompanying drawings and detailed description.
Fig. 1 is the naval vessel confirmation side described herein based on the naval vessel plane geometry characteristic of visible remote sensing image on star
Method flow diagram;
Fig. 2 is the comparison for carrying out accurately approaching the result figure and original image on potential target boundary using context of methods;
Fig. 3 be under different resolution in a variety of marine backgrounds pending naval vessel confirmation visible remote sensing image slice and
The naval vessel of context of methods confirms result figure.
Specific embodiment
In the following, true to the naval vessel provided herein based on the naval vessel plane geometry characteristic of visible remote sensing image on star
Verifying method specific implementation step is described in detail, and process flow is shown in Fig. 1.
Step 1: multiple ship images to be detected being extracted to the visible remote sensing image of input and are sliced, wherein each
It only include a potential Ship Target in ship images slice to be detected.
Step 2: ship images to be detected slice being traversed, finds out its mean value and standard deviation, wherein mean value and variance
Calculation formula it is as follows:
Wherein μ is mean value, and σ is standard deviation, and f (x, y) is to be sliced the gray value that upper coordinate is (x, y), and w and h are respectively represented
The width and height of the slice;Karma gram fast square root algorithm is used when radical sign operation is opened in calculating.Wherein, f (x, y) and f (x,
y)2Summation operation calculated in same circulation, the quadrature operation of wide w and high h of ship images to be detected slice is also only counted
It calculates primary.When carrying out binarization operation, the calculation formula used is as follows:
G'(x, y)=l { f (x, y) > μ+σ } (3)
Wherein g'(x, y) it is the value that (x, y) is put after binaryzation, l { } is indicator function, value rule are as follows: { value is l
Genuine expression formula }=1;L { value is false expression formula }=0.Provide potential target T'{ (x, y) | g'(x, y)=1.It calculates potential
The mass center m of Ship Target T'0Position (x0,y0):
It is wherein all to be related to g'(x, y) summation operation all carried out in same circulation.
Step 3: traversing binarization result using 8 neighborhoods, and find mass center m0' ∈ T' region;Traverse naval vessel to be detected
Slice calculates probability of occurrence p of the different gray values in entire slicei:
Wherein h (i)x,yGray value is represented on the slice as the indicator function of each pixel of i;
h(i)x,y=l { f (x, y)=i } (6)
Wherein f (x, y)=i represents the value of pixel (x, y) in the slice as i;Calculate each pixel and m0' point it is opposite
Entropy:
Formula (6) calculates any point Q and center of mass point m0' between relative entropy,For point m0Gray value appearance it is general
Rate, pQFor the probability of occurrence of the gray value of point Q;Wherein Logarithmic calculation is bottom quickly to calculate with 2.Then according to relative entropy
Calculated result it is using K-means algorithm that relative entropy is lesser with mass center m0' gather for one kind, and by these in binary map g'
Point is set to 1 and shows, and obtain update potential Ship Target T (x, y) | can be with mass center m0Carry out 8 neighborhood connections } two
Value figure g and the area S for recording potential Ship Target TT, m is updated with formula (4)0Position.
Step 4: calculating the length and width ratio r and mass center m of the slice0To the distance of centre of slice point:
R=w/h (8)
dx=| x0-w/2| (9)
dy=| y0-h/2| (10)
Wherein dxFor mass center m0With centre of slice point in the direction of the x axis at a distance from, dyFor existing with centre of slice point for mass center
Distance on y-axis direction.Following three kinds of situations will be divided into according to the calculated result of formula (8), (9) and (10) and find out mass center m1Position
It sets:
(I) if 2/3 < r < 3/2 and dx< 2, dy< 2, then first taken from the binary map g of the slice sub-block s (x, y) | x
< w/2, y < h/2 } and its mass center m is found out using formula (11)1Position (x1,y1):
Mass center m is calculated using formula (12)1Position (x1,y1) arrive s central point distance:
sdx=| x1-w/4|
sdy=| y1-h/4| (12)
Wherein sdxFor mass center m1With sub-block s central point in the direction of the x axis at a distance from, sdyFor mass center and sub-block s central point
Distance in the y-axis direction;If sdx>=2 or sdy>=2, then taken from binary map g again sub-block s (x, y) | x < w/2, h/2≤
Y≤h } and its mass center m is sought with formula (11)1。
(II) if r > 3/2, s { (x, y) | x < w/2 } is taken from binary map g as sub-block and seeks its mass center with formula (11)
m1。
(III) if r < 2/3, s { (x, y) | y < h/2 } is taken from binary map g as sub-block and seeks its matter with formula (11)
Heart m1。
Step 5: using m obtained in step 2 and step 30With m1Point following two situation calculates potential Ship Target T's
Symmetry axis l:Ax+By+C=0.
(I) if | x0-x1| > 0.1 calculates symmetry axis using following formula:
A=-k, B=1, C=-b (15)
(II) in addition to meeting (I) the case where, calculates symmetry axis using following formula:
A=1, B=0, C=-x0 (16)
Step 6: according to the slope of calculated symmetry axis l, point following two situation calculating is located at symmetry axis both sides
Apart from farthest point e1And e2And this two o'clock is to m0Distance ed1And ed2:
(I) if B=0, with x0Centered on along y-axis respectively to the left side and the right traversal find point e1And e2And it records respectively
Its distance ed1And ed2, and calculate with formula (17) distance that traversed point (x, y) arrives symmetry axis:
D (x, y)=| y-y0| (17)
(II) if B ≠ 0 and, calculated m with formula (18)0It puts and perpendicular to the linear equation l' of symmetry axis:
And the point for belonging to potential Ship Target T in e { (x, y) | A'x+B'y+C'< 1,1≤x≤w, 1≤y≤h } is found,
With m0Centered on, point e is found out with formula (19) to both ends1And e2And ed1And ed2:
The point for belonging to potential Ship Target T in e { (x, y) | Ax+By+C < 1,1≤x≤w, 1≤y≤h } is found out, and with m0
It is found out in e ∩ T using formula (19) apart from center of mass point m to upper and lower respectively for starting point along symmetry axis0Farthest point v1And v2And
This two o'clock is to m0Distance vd1And vd2。
Step 7: finding out the area M and ratio r of the minimum circumscribed rectangle of potential Ship Target Tls:
M=(ed1+ed2)×(vd1+vd2) (20)
Area S according to the area M and potential Ship Target TTRatio and rlsValue make following judgement:
(I) if obtaining 1 < M/ST< thre1 and rls> thre2 then determines that potential Ship Target T is naval vessel;
(II) determine that potential Ship Target T is not naval vessel if not meeting (I).
Wherein, thre1 and thre2 is preferably 1.2 and 2 with the relationship of resolution ratio according to true naval vessel.
The application has carried out algorithm experimental on a pc platform, and experimentation has used 1 meter, 2 meters, 5 meters resolution ratio respectively
The naval vessel undetermined slice extracted in visible remote sensing image is 20240 total, and every image slice is not of uniform size, wherein really
Slice comprising naval vessel is 5060.Fig. 2 gives the ship images undetermined slice and this hair used during part Experiment
The bright potential target result accurately extracted in naval vessel slice undetermined of method.Fig. 3 gives the method for the present invention to naval vessel figure undetermined
As the judgement result of potential target in slice.It can be seen from the figure that the application method can be directed under different rates respectively from a variety of
Potential target and background are effectively separated in marine background;The method of the present invention can be rapidly from the non-naval vessel such as wave, cloud, harbour
Ship Target is separated in object.
Context of methods is compared with the naval vessel confirmation method that existing visible light naval vessel detects.Method in comparative experiments
1. for Changren Zhu et al. in 2010 in " IEEE Transactions on Geoscience and Remote
Sensing " " the A Novel Hierarchical Method of Ship Detection from that delivers of periodical
The local proposed in Spaceborne Optical Image Based on Shape and Texture Features " text
multiple patterns(LMP);Method be 2. Fukun Bi et al. in 2012 in " IEEE Geoscience and
Remote Sensing Letter " " the A Visual Search Inspired Computational Model that delivers of periodical
The local context proposed in for Ship Detection in Optical Satellite Images " text
facilitation(LCF);Method be 3. Gong Cheng et al. in 2016 in " IEEE Transactions on
Geoscience and Remote Sensing " " the Learning Rotation-Invariant that delivers of periodical
Convolutional Neural Networks for Object Detection in VHR Optical Remote
Rotation-invariant CNN (RICNN) method proposed in Sensing Images " text.
The application is compared in objective indicator with the above method using accuracy, False Rate, and calculation method is as follows:
The comparing result of the present processes and the above method is as shown in table 1.The application method and existing as can be seen from the table
Have disclosed method compare have can reach higher timeliness and accuracy while guaranteeing low False Rate.
1 remote sensing images naval vessel confirmation method experimental result of table
It is true to the naval vessel provided herein based on the naval vessel plane geometry characteristic of visible remote sensing image on star above
Verifying method is described in detail, it is apparent that the specific implementation form of the application is not limited thereto.For the art
For those skilled in the art, it is carried out without departing substantially from claims hereof range various obvious
Change all within the scope of protection of this application.
Claims (6)
1. visible remote sensing image Ship Detection on the star based on geometrical characteristic, it is characterised in that include the following steps:
(1) it extracts multiple ship images to be detected to the visible remote sensing image of input to be sliced, wherein each warship to be detected
It only include a potential Ship Target in ship image slice;
(2) ship images to be detected slice is traversed, finds out the mean value and standard deviation of grey scale pixel value in the slice, and
It is sliced the binary map g' for carrying out binarization operation to obtain the slice to described image, is acquired currently according to the binary map g'
The mass center m for the potential Ship Target T' for including in slice0' position;
(3) using obtained mass center m in above-mentioned steps (2)0' position and binary map g', search mass center m0' belonging to connected region
Domain and according to each pixel and mass center m0' relative entropy as a result, accurately approach the boundary in the region by K-means algorithm,
And updated binary map g is obtained, potential Ship Target T and its mass center m0;
(4) the length-width ratio r and the mass center m of the slice are calculated0It is special according to shipform to the distance of the centre of slice point
Property the sub-block with symmetric relation is obtained from ship images to be detected slice, and calculate the mass center m of the sub-block1Position;
(5) according to above-mentioned two mass center m0And m1Position, calculate the symmetry axis straight line of potential Ship Target T, and obtain potential
Locality of the Ship Target T in the slice;
(6) with the mass center m in the slice0For starting point, the symmetry axis straight line that foundation above-mentioned steps (5) obtains is respectively symmetrical
Axis straight line two sides, which are found, belongs to potential Ship Target T, and the point e farthest apart from symmetry axis vertical range1And e2, and find described
The point v that symmetry axis intersects with potential Ship Target T external periphery outline1And v2;
(7) the area M of potential Ship Target T minimum circumscribed rectangle, and the point e according to obtained in step (6) are calculated1, e2, v1And v2
Each point is calculated separately to mass center m0Distance ed1, ed2, vd1And vd1, according to described apart from ratio calculatedIf
The area M and ratio rlsMeet default Rule of judgment, then determines that potential Ship Target T is naval vessel, otherwise determine potential warship
Ship target T is not naval vessel.
2. method as described in claim 1, it is characterised in that:
In the step (2), when carrying out binarization operation, the calculation formula used is as follows:
G'(x, y)=l { f (x, y) > μ+σ } (1)
Wherein μ is mean value, and σ is standard deviation, and f (x, y) is to be sliced the gray value that upper coordinate is pixel (x, y), g'(x, y) it is two
The value that (x, y) is put after value, l { } are indicator function, value rule are as follows: l { value is genuine expression formula }=1;{ value is false to l
Expression formula=0;Then the mass center m of potential Ship Target T is calculated using following formula (2)0' position (x0,y0):
Wherein g'(x, y) be binaryzation after pixel (x, y) value.
3. method as described in claim 1, it is characterised in that:
In the step (3), different gray values going out in entire slice is calculated first with formula (3) to the calculating of relative entropy
Existing Probability pi:
Wherein h (i)x,yGray value is represented on the slice as the indicator function of each pixel of i;
h(i)x,y=l { f (x, y)=i } (4)
Wherein f (x, y)=i represents the value of pixel (x, y) in the slice as i;Then the slice is calculated according to formula (5)
Middle any pixel point Q and mass center m0The relative entropy of point:
Wherein,For point m0Gray value probability of occurrence, pQFor the probability of occurrence of the gray value of point Q.
4. method as described in claim 1, it is characterised in that:
In the step (4), the operation of sub-block is extracted from the slice and calculates mass center m1The process of position is cut according to described in
The length and width ratio r of piece is divided into following three kinds of situations:
(I) if 2/3 < r < 3/2 and dx< 2, dy< 2, then first extracted from the binary map g of the slice sub-block s (x, y) | x <
W/2, y < h/2 } and its mass center m is found out using formula (6)1Position (x1,y1):
And the mass center m is calculated according to formula (7)1To the distance of s central point:
sdx=| x1-w/4|
sdy=| y1-h/4| (7)
Wherein sdxFor mass center m1With sub-block s central point in the direction of the x axis at a distance from, sdyFor mass center with sub-block s central point in y
Distance in axis direction;If sdx>=2 or sdy>=2, then taken from binary map g again sub-block s (x, y) | x < w/2, h/2≤y≤
H } and its mass center m is sought using formula (6)1;
(II) if r > 3/2, s { (x, y) | x < w/2 } is taken from the binary map g as sub-block and seeks its mass center with formula (6)
m1;
(III) if r < 2/3, s { (x, y) | y < h/2 } is taken from the binary map g as sub-block and seeks its mass center with formula (6)
m1。
5. method as described in claim 1, it is characterised in that:
In the step (7), the area M of Ship Target T minimum circumscribed rectangle is calculated using formula (15):
M=(ed1+ed2)×(vd1+vd2) (15)
Wherein ed1, ed2, vd1And vd1Respectively point e1, e2, v1And v2To m0Distance, using formula (16) ratio calculated rls:
6. method as described in claim 1, it is characterised in that:
Area S in the step (7), according to potential Ship Target TTWith the area of the minimum circumscribed rectangle of the Ship Target T
The ratio and r of MlsValue make following judgement:
(I) if obtaining 1 < M/ST< thre1 and rls> thre2 then determines that potential Ship Target T is naval vessel;
(II) determine that potential Ship Target T is not naval vessel if not meeting (I);
Wherein, thre1 and thre2 is preferably 1.2 and 2 with the relationship of resolution ratio according to true naval vessel.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710421783.4A CN109002747A (en) | 2017-06-07 | 2017-06-07 | Visible remote sensing image Ship Detection on star based on geometrical characteristic |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710421783.4A CN109002747A (en) | 2017-06-07 | 2017-06-07 | Visible remote sensing image Ship Detection on star based on geometrical characteristic |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109002747A true CN109002747A (en) | 2018-12-14 |
Family
ID=64573062
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710421783.4A Pending CN109002747A (en) | 2017-06-07 | 2017-06-07 | Visible remote sensing image Ship Detection on star based on geometrical characteristic |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109002747A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110287817A (en) * | 2019-06-05 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Target identification and the training method of Model of Target Recognition, device and electronic equipment |
CN112381838A (en) * | 2020-11-14 | 2021-02-19 | 四川大学华西医院 | Automatic image cutting method for digital pathological section image |
CN118052997A (en) * | 2024-04-16 | 2024-05-17 | 北京航空航天大学 | Target confirmation method embedded with physical characteristics and common sense |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105574488A (en) * | 2015-12-07 | 2016-05-11 | 北京航空航天大学 | Low-altitude aerial infrared image based pedestrian detection method |
-
2017
- 2017-06-07 CN CN201710421783.4A patent/CN109002747A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105574488A (en) * | 2015-12-07 | 2016-05-11 | 北京航空航天大学 | Low-altitude aerial infrared image based pedestrian detection method |
Non-Patent Citations (5)
Title |
---|
FENG YANG等: "SHIP DETECTION FROM OPTICAL SATELLITE IMAGES BASED ON VISUAL SEARCH MECHANISM", 《IGARSS 2015》 * |
LI ZHIGUO等: "Warship detection in Smoke Screen Interference Based on Fractal and Rough Set Theory", 《ICSP2012 PROCEEDINGS》 * |
王方超等: "基于几何特征的水面舰船快速检测算法", 《海军工程大学学报》 * |
程小青等: "基于对称轴的自然图像中物体部件检测", 《中国科技论文》 * |
陈文婷等: "SAR 图像舰船目标识别综述", 《现代雷达》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110287817A (en) * | 2019-06-05 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Target identification and the training method of Model of Target Recognition, device and electronic equipment |
CN112381838A (en) * | 2020-11-14 | 2021-02-19 | 四川大学华西医院 | Automatic image cutting method for digital pathological section image |
CN112381838B (en) * | 2020-11-14 | 2022-04-19 | 四川大学华西医院 | Automatic image cutting method for digital pathological section image |
CN118052997A (en) * | 2024-04-16 | 2024-05-17 | 北京航空航天大学 | Target confirmation method embedded with physical characteristics and common sense |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106022288B (en) | The identification of marine oil spill information and extracting method based on SAR image | |
CN104217196B (en) | A kind of remote sensing image circle oil tank automatic testing method | |
CN110414411A (en) | The sea ship candidate region detection method of view-based access control model conspicuousness | |
CN109427055B (en) | Remote sensing image sea surface ship detection method based on visual attention mechanism and information entropy | |
CN104361582B (en) | Method of detecting flood disaster changes through object-level high-resolution SAR (synthetic aperture radar) images | |
CN103886285A (en) | Method for detecting optical remote-sensing image ship with assistance of prior geographic information | |
CN107392885A (en) | A kind of method for detecting infrared puniness target of view-based access control model contrast mechanism | |
Rau et al. | Bridge crack detection using multi-rotary UAV and object-base image analysis | |
CN108121991A (en) | A kind of deep learning Ship Target Detection method based on the extraction of edge candidate region | |
CN103020975A (en) | Wharf and ship segmentation method combining multi-source remote sensing image characteristics | |
CN106651880B (en) | Offshore moving target detection method based on multi-feature fusion thermal infrared remote sensing image | |
CN113963222B (en) | High-resolution remote sensing image change detection method based on multi-strategy combination | |
CN107704865A (en) | Fleet Targets Detection based on the extraction of structure forest edge candidate region | |
CN108830224A (en) | A kind of high-resolution remote sensing image Ship Target Detection method based on deep learning | |
CN109002747A (en) | Visible remote sensing image Ship Detection on star based on geometrical characteristic | |
CN109063669B (en) | Bridge area ship navigation situation analysis method and device based on image recognition | |
CN107862262A (en) | A kind of quick visible images Ship Detection suitable for high altitude surveillance | |
Cai et al. | Broken ice circumferential crack estimation via image techniques | |
Zhang et al. | Nearshore vessel detection based on Scene-mask R-CNN in remote sensing image | |
CN114821358A (en) | Optical remote sensing image marine ship target extraction and identification method | |
Yasir et al. | ShipGeoNet: SAR image-based geometric feature extraction of ships using convolutional neural networks | |
CN116703895A (en) | Small sample 3D visual detection method and system based on generation countermeasure network | |
Zhao et al. | A robust lane detection algorithm based on differential excitation | |
Wang et al. | A multidirectional one-dimensional scanning method for harbor detection from sar images | |
Shen et al. | Polarization calculation and underwater target detection inspired by biological visual imaging |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181214 |